October 2000 Newsletter - Naval Postgraduate School

PROJECT NOTES
SYSTEMATIC APPROACH TO TROPICAL CYCLONE TRACK FORECASTING
Research Associate Professor Lester E. Carr, III
Distinguished Professor Russell L. Elsberry
Department of Meteorology
Introduction
Tropical cyclones are a major threat to Fleet units afloat and
bases ashore. The Joint Typhoon Warning Center in Pearl
Harbor, Hawaii, is responsible for providing DoD units
warnings of tropical cyclones in the western North Pacific and
in the Southern Hemisphere to the east coast of Africa. This
center, which moved from Guam in January 1999 due to a
base closure action, has re-started in Pearl Harbor with a
nearly 100% turnover in personnel.
The meteorology groups in Monterey provide essential
support for the Joint Typhoon Warning Center. The Fleet
Numerical Meteorology and Oceanography Center provides
numerical analyses and forecasts as well as some satellite data.
The Naval Research Lab – Monterey has developed a number
of those forecast products, and also the Automated Tropical
Cyclone Forecast System that assists the forecaster in data
handling, visualization, and warning preparation. The
Department of Meteorology faculty research and theses have
contributed many of the scientific bases for new products.
A long-term research project entitled the “Systematic
Approach to Tropical Cyclone Track Forecasting,” led by the
authors has come to fruition this summer with an operational
test at the Joint Typhoon Warning Center. Basic research in
the understanding of tropical cyclone motion sponsored by
the Office of Naval Research (ONR) began more than a
decade ago. A new ONR applied research project to incorporate
these new understandings in a knowledge-based expert
system was begun in 1996. More recently, the Space and
Naval Warfare Command has provided funding to transition
the research to operations at the Joint Typhoon Warning
Center.
Research Goals
The long-range objective is to provide a product that will
assist the forecaster in reducing tropical cyclone track forecast
errors at 72 hours to the Joint Chiefs of Staff goal of 150
nautical miles. The primary guidance used by the forecaster
to make these forecasts in the western North Pacific come
from international numerical weather forecast centers: a
global and a regional model from Fleet Numerical Meteorology
and Oceanography Center; global and regional model
from the Japan Meteorological Agency; and a global model
from the United Kingdom Meteorological Office. Since
1994, these dynamical models have been markedly improved
for tropical cyclone track forecasting. However, the
model track forecasts still occasionally have large errors,
sometimes as large as 1000 nautical miles after 72 hours.
What is needed is guidance on when to accept or to reject
the model track guidance for a particular tropical cyclone.
Thus, the short-term goal was to develop an expert system
module that would assist the forecaster in rejecting dynamical
model track guidance that is likely to be in error by
more than 300 nautical miles at 72 hours. The result of
rejecting that erroneous model track should lead to a more
accurate warning than if all five model tracks were simply
averaged.
Dynamical Model Traits Knowledge Base
An exhaustive study was made of all large track errors made
by the two Navy models during the 1997 tropical cyclone
season. Approximately one third of all 72-hour forecasts
have these large errors. Seven frequently occurring trackerror
mechanisms were discovered. Conceptual models
were developed that describe how the error mechanisms
appear in the wind and/or sea-level pressure forecast fields.
Comprehensive case studies that illustrate the evolution of
these wind or pressure fields over 72 hours were developed
for the guidance of forecasters. This information was
collected in a NPS technical report issued in June 1999.
Subsequent studies, including a M.S. thesis by LT Dave
Brown, USN, have demonstrated that these same error
mechanisms apply for large track errors by other dynamical
models and in other tropical cyclone basins.
Expert System Module
Whereas the above dynamical model traits knowledge base
was retrospective, a tool was needed to assist the forecaster
in recognizing in real-time that a track forecast was likely to
be erroneous. Programming the expert system module
turned out to be a formidable task because the information
has to be provided in a logical manner, be rapidly available,
and be “forecaster-friendly.” Jim Peak of the Computer
Sciences Corporation, who previously worked for the Naval
Postgraduate School, did most of the expert system coding.
The Automated Tropical Cyclone Forecast System developed
by the NRL-Monterey was helpful in providing the
necessary data streams.
--continued on page 21
NPS Research page 19
October 2000